This package provides a python framework for building and testing machine learning models for time series and tabular data.
What's Changed
* updated master with dev by AtrCheema in https://github.com/AtrCheema/AI4Water/pull/12
* Dev by AtrCheema in https://github.com/AtrCheema/AI4Water/pull/13
**Full Changelog**: https://github.com/AtrCheema/AI4Water/compare/v1.0-beta...v1.0-beta.1
- predict method returns only predicted array by default. If the user wants true array as well, return_true should be set to True.
- changed name of sub-modules pre_processing => preprocessing, post_processing => postprocessing, hyper_opt => hyperopt and ETUtil => et
- improved docs
- removed bugs
- transformations classes have config and from_config methods
- integration of nbeats
- added fit_with_tpot method for experiments
- added autocorrelation and partial autocorrelation in eda
- predict method does not receive prefix method
- added examples
- datahandler can read xlsx, csv, mat, npz, netcdf pqrquet and feather file types.
- added regplot function as part of utils
- predict method can calculate variety of errors. By default now calculates minimal errors
- improved speed of tests and made them less verbose
- unified ShapExplainer class for both ml and dl models
- unified LimeExplainer class for both ml and dl models
- Experiment class can handle verbosity argument better
- improved interdependency of packages in different sub-modules i.e. shapfile not required if datasets sub-module is not used
- pin the versions with which ai4water is tested
- predict can take user defined arguments just as keras model or sklearn model
- added conditionalize layer which is part of ConditionalRNN